Two new datasets point the same direction: the early signals of cognitive decline are hiding in routine physiology—if you know where to look.
A report from News-Medical highlights a finding with immediate clinical relevance: daytime cortisol patterns correlate with trajectories of cognitive decline, yet show no predictive link to Alzheimer's disease specifically. The distinction matters. It separates general neurocognitive erosion—attentional lapses, processing-speed drops, executive-function drag—from the amyloid-driven pathology of Alzheimer's, and it implies that salivary or serum cortisol mapping could become a low-cost screening layer for at-risk populations before formal neurocognitive testing kicks in. Without the underlying study text available, the mechanism remains unspecified here, but the headline signal alone reframes cortisol from a blunt stress marker to a potentially granular cognitive-performance index.
SleepFM: AI recovers what clinicians discard
A second data point reinforces the same logic. Researchers involved in a Cleveland Clinic–IBM collaboration developed SleepFM, an AI model trained on standard polysomnography data from the clinic's STARLIT registry and validated against a nationwide cohort. The model identified five distinct patient risk subtypes with sharply divergent long-term health trajectories—subtypes invisible to the conventional apnea-hypopnea index. Patients in the highest-risk category faced double the five-year mortality risk compared with the lowest-risk group. Critically, the model surfaced hidden signal patterns linked to heart disease, cognitive decline, and death from the same overnight recordings clinicians have been performing for decades.
The key statistic: an estimated one to four million polysomnograms are conducted annually in the United States alone. Historically, clinicians have distilled each into a handful of summary measures. SleepFM suggests the remaining data contains clinically meaningful physiological signatures that current practice simply throws away.
Why this matters for cognitive performance tracking
Together, these two signals converge on a single operational principle: the early markers of cognitive decline are already embedded in routine, repeatable physiological measurements—cortisol rhythms during waking hours, electrophysiological patterns during sleep. Neither requires exotic biomarkers or expensive imaging. Both require better extraction.
For practitioners optimizing cognitive performance in non-pathological populations, the implication is straightforward. Periodic daytime cortisol profiling—morning, midday, evening samples at minimum—may serve as a leading indicator of attentional and executive-function drift, even when Alzheimer's-specific pathology is absent. Pairing that with longitudinal sleep-study data, ideally processed through models that capture full-signal richness rather than summary indices, could provide a composite cognitive-risk dashboard far earlier than conventional neuropsych batteries detect deficit.
What to watch: whether the cortisol–cognitive-decline association holds across age cohorts and whether SleepFM-class models move from retrospective registry analysis to real-time clinical integration. Until then, the protocol is pragmatic—track cortisol rhythmicity, commission granular sleep-data reviews, and treat both as performance baselines rather than diagnostic endpoints.